Proposed statistical analysis to evaluate qualitative proficiency testing of Salmonella serotyping

Abstract

Salmonella is one of the main causes of foodborne illnesses in humans worldwide. More than 2500 different serotypes have been described so far. Correct identification of the serotype is essential, since the identification of different serotypes in animal populations and foodstuffs can lead to the application of different corrective measures according to current European food and animal legislation. Since 2007, Istituto Zooprofilattico Sperimentale delle Venezie, in its capacity as Italian National Reference Laboratory for Salmonella, has organised one proficiency test (PT) for Salmonella serotyping per year. The performance of the participant laboratories is assessed based on their capability to correctly identify serotypes of 20 selected Salmonella strains. To evaluate the overall PT performance, a detailed descriptive analysis of all results (identification of serotype by somatic and flagellar antigens for each strain) has been performed. Moreover, the performance of each participant is assessed using the generalised Cohen’s κ statistic for multiple categories. The aim of the present study is to describe the Salmonella serotyping PT, focusing on the statistical approach followed to analyse the data collected. As an example, analysis of data obtained during the last PT (2013) is reported. All participants (12 laboratories) obtained κ values indicative of substantial or almost perfect agreement with the expected results. The overall performance of the PT also indicated perfect agreement among participants (κ value equal to 0.8767). Moreover, from 2007 to 2013, the κ value over time was always significantly higher than 0.75, indicating a substantial agreement among all participants in the qualitative PT for Salmonella serotyping.